现代妇产科进展2025，Vol.34Issue(8)：588-593,6.DOI:10.13283/j.cnki.xdfckjz.2025.08.003
基于机器学习算法的LSIL患者漏诊HSIL+的模型构建与验证
The construction and validation of a machine learning-based model for predicting missed diagnosis of HSIL+in LSIL patients
摘要
Abstract
Objective:To investigate the risk factors for missed diagnosis of high-grade squamous intraepithelial lesions(HSIL+)in patients with low-grade squamous intraepithelial le-sions(LSIL)and to develop a predictive model.Methods:A retrospective analysis was conduc-ted on LSIL patients who underwent loop electrosurgical excision procedure(LEEP)at Qilu Hospital of Shandong University from January 2017 to January 2024.Clinical data,including age,menopausal status,TCT results,HPV16/18 status,cervical transformation zone type,and colposcopic impressions,were collected.Important variables were identified using Lasso regres-sion,and a predictive model was constructed through multivariate logistic regression.The model's discrimination,calibration,and clinical applicability were evaluated using receiver oper-ating characteristic(ROC)curves,calibration curves,decision curve analysis(DCA).Results:Lasso regression identified HPV16/18 positivity,TCT≥ASC-H,cervical transformation zone type 3,and colposcopic impression of high-grade lesions as significant risk factors for missed di-agnosis of HSIL+.The predictive model demonstrated an area under the curve(AUC)of 0.92 in the training set,with a sensitivity of 0.85 and specificity of 0.88.In the validation set,the AUC was 0.82,with a sensitivity of 0.83 and specificity of 0.80,indicating good predictive performance.Calibration curves and the Hosmer-Lemeshow test showed good model fit(training set x2=3.50,P=0.61;validation set x2=2.89,P=0.62).DCA curves demonstrated that the model had good clinical applicability.Conclusion:HPV16/18 positivity,TCT≥ASC-H,cervical type 3 transformation zone,and colposcopic impression of high-grade lesions are significant risk factors for the missed diagnosis of HSIL+in LSIL patients.The predictive model exhibits good discrimination and clinical applicability,providing valuable guidance for individualized risk as-sessment and treatment strategies in clinical practice.关键词
低级别鳞状上皮内病变(LSIL)/漏诊/高级别鳞状上皮内病变(HSIL+)/预测模型/宫颈癌Key words
Low-grade squamous intraepithelial lesion(LSIL)/Missed diagnosis/High-grade squamous intraepithelial lesion(HSIL+)/Predictive model/Cervical cancer分类
医药卫生引用本文复制引用
李佳轩,王志玲,杨兴升..基于机器学习算法的LSIL患者漏诊HSIL+的模型构建与验证[J].现代妇产科进展,2025,34(8):588-593,6.基金项目
国家重点研发计划(No:2023YFC2705805) （No:2023YFC2705805）
